During the last three decades, increased attention has been devoted towards psychological variables influencing injury risk (Hackfort & Kleinert, 2007). Of these prediction studies, a majority have used prospective designs with one single measurement point and continuous injury recording over a number of weeks. In order to grasp the changes in those variables, the use of repeated measure designs with multiple measurement points is warranted. Obtaining data from multiple points will enable use of advanced statistics, such as latent growth curve analysis. Unlike regular analyses (e.g., ANOVAs), growth curve analyses focus on within-person change and how within-person changes in state variables could affect injury risk. Based on findings from injury prediction research, investigators have targeted such variables (e.g., daily hassles, coping) in experimental studies aimed at preventing injuries. A meta-analysis, covering seven experimental studies, showed most studies to be effective in decreasing the number of injuries in the experimental groups (overall Hedges g Effect size = .81; Tranaeus, Ivarsson & Johnson, submitted). Even if the experimental studies have used true or quasi-experimental designs, several methodological issues can be addressed. First, in most of the studies a number of different mental skills are included in the intervention approach leading to difficulties in differentiating which specific mental skills may be responsible for producing reductions in injury. Second, since most of the experimental studies conducted used no-attentional control groups (i.e., the participants in these groups will not be given a placebo treatment), it is likely that large effects could be explained by the Hawthorn effect. Third, in most studies, researchers discuss the importance of their results based on suggested cut-off criteria for the p-values and/or effect sizes (ES). This procedure could be addressed as a limitation since p-values and/or effect sizes do not indicate anything about the results’ clinical significance (e.g., Ivarsson, Andersen, Johnson & Lindwall, 2013). Also, the fact that non-adjusted ES, which were reported in all studies providing ES, are positively biased due to sampling error (Synder & Lawson, 1993) might have led to overestimation of the intervention effects. This presentation will (a) highlight the designs of previous prediction studies while focusing on advantages of longitudinal repeated-measure designs (b) discuss different experimental designs that have been used in injury prevention research and, (c) suggest methodological and statistical considerations for future research on injury prevention.